mainyehc

rcfa said:

The real issue isn’t if M$ is going to port Windows10-ARM in some secret handshake deal with Apple, the key question is, will Apple publicly document their hardware well enough that anyone can port any OS to it.

Like running Linux, FreeBSD, bare metal virtualization software allowing near-instant switching between macOS and other operating systems, etc. etc.

Someone should have asked Craig Federighi THAT question...

What are you talking about? Apple, specifically Craig Federighi, directly addressed that in more than one occasion. M1 chips support that hypervisor framework that obviates the usage of kernel extensions, and a Linux VM was even demoed during the WWDC keynote.

Of course it’s all properly documented, and judging from Parallels’ development blog, Apple even partnered behind the scenes with them. Apple’s recent comments, if you know how to do the Kremlinology that comes with following the company, tell you everything you need to know: it’s up to Microsoft to offer full, non-OEM versions of ARM64 Windows.

Maybe it’s not just a licensing issue, as the M1 is vastly different from those puny Qualcomm offerings, so maybe there’s more work involved. But after the basics are covered, Parallels, VMWare, Oracle, the OSS community, etc., only have to bridge the gap. It’s not like Microsoft has to develop a “VirtualPC redux”.

Interestingly, that developer said in his Twitter that x86-32 emulation was decent, so maybe Microsoft’s equivalent to Rosetta 2 isn’t *that* crappy; its abysmal performance is maybe due to those ARM PCs being severely underperforming.

I’ll just throw in my €0,02: what if the Unified Memory Architecture is the key here?

No, really, hear me out: what do conventional PCs (even PowerPC Macs back in the day) do when they run out of memory? They do paging of stuff from memory into storage, A.K.A. virtual memory.

The thing is, I’m guessing the on-chip memory of these things is not only more power efficient, but fast. Really fast. Maybe even much faster than conventional memory in DIMM and SO-DIMM modules connected to the CPU via its memory controller, as it resides on the SoC itself and benefits from direct, fast lanes connecting it to the CPU cores themselves.

What about SSDs? Well, those are also really, really fast, even for such a paltry, apparently non-Pro machine like this. While they’re not exactly melded into the CPU as well, it seems Apple has done some advances in this area as well.

So… maybe paging to virtual memory isn’t as much of an issue as it was before? And maybe said UMA is better at managing the whole process as well?

If Apple adds a “Mac mini Pro”, or “Mac Pro mini”, or whatever they call it to the lineup, in space gray to match the later Intel models and the DTK, with an M1X/M2 chip, four TB4/USB4 ports, and, say, 32 GB of this newfangled UMA memory, expect smaller studios to pick this thing up. It’s the trascan-redux-as-lunchbox, minus the melting and inefficient dual GPUs and at A THIRD of the price, and it might very well sell like hotcakes.

For now, this first generation is more of a proof of concept than anything else (at least as far as some less demanding segments of the professional market are concerned), but keep an open mind and your eyes peeled for benchmarks and real-world testing.

Hank2.0 said:

Considering the ever increasing computational power being put into iPhones, I'm surprised Apple hasn't played with the idea of a processor-less iMac or MacBook where you would slide your iPhone into slot and it would act as the processor, if you get my meaning.

Why would they do that (at least branded as an “Mac”, and not as a glorified display/hub)? Apple wants you to buy as many fully-fledged, standalone devices as possible, and they reward you accordingly with whatever performance their TDP and thermal envelope allows.

What a stupid, stupid prototype… Great idea (no, really), terrible execution (and not because it looks ugly, but because it was a dead end, as it was so very well put in the article).

ArianneFeldry said:

It's taken about 20 years for Unreal to be the juggernaut it is today. That you think Apple can just walk in and make something to compete with it in a year or two is laughable. What, are they going to put more money into Metal? The failed API that was supposed to invigorate the gaming industry on Apple devices? Meanwhile gaming on MacOS has been reversing with the lack of 32-bit support and the ARM transition has killed VR development on Mac's with SteamVR no longer being updated for the platform. 
Throwing money at something does not equate to success. If it did we would be using Windows Phones and Cortana would be the go-to voice assistant. And Rosetta 2 running faster than native code? I highly doubt that. 
Stop exaggerating Apple's accomplishment with market valuation, and stop underestimating Unreal.

Who’s to say Apple doesn’t have an internal game engine prototype already laying around in some secret lab?

As for Rosetta 2 running translated x86 code faster than an equivalent Intel-based Mac could do natively, don’t count that out… Don’t forget that those demos at WWDC were made on an iPad Pro SoC, not on a Mac-specific design.

Also, if you’ve been an Apple user long enough, you should remember how MacBooks, Mac Pros and Intel iMacs and Mac minis were in some cases faster at running PowerPC code translated on the fly with Rosetta 1 than their G4 and G5 counterparts were natively. No one is arguing that Rosetta 2-translated binaries (which, mind you, undergo that process during installation, and not at runtime) run as fast as ARM binaries on Apple Silicon…